Analog AI Computing
With Moore’s law reaching its end and Dennard’s scaling already hitting the wall, digital accelerators will soon not be enough to execute demanding workloads efficiently. These accelerators are highly power and area inefficient.
This research proposes a truly scalable and reconfigurable near-memory analog accelerator chip for Energy-Efficient Computing.
Scalable Quantum Control and Readout System (SQ-CARS) is designed to be scalable, configurable, and phase-synchronized, providing multiqubit control and readout capabilities.
The next generation computational architectures require embedding deep learning, machine learning and artificial intelligence onto computationally efficient hardware and devices with low energy footprints. This is necessitated by an ever-increasing demand in computing power, which, coupled with data explosion is pushing the boundaries of conventional computational architectures ······
The neuromorphic architectures are inspired by neurobiological systems and have characteristics such as in-memory computing, stochastic computing and massive network fan-outs. They are parallel and distributed architectures which exhibit low precision along with adaptive learning ·····
High energy-efficiency and low computational/ memory foot-print are the key design requirements due to limited battery resources. In this regard, wake-up systems play an integral role and operate by triggering on the computationally and power-intensive modules only when some ambient conditions are detected. Memtransistor crossbar arrays have shown great potential for neuromorphic computing and learning ·····
These non-volatile memories if arranged in crossbar pattern as shown can carry out array size in-memory MVM and addition using Kirchhoff’s current law (KCL) in a single time step. Each column of memtransistor crossbar array represents single template vector with rows equal to the number of features stored as memductance (conductance of memtransistor device) ·····
The neurobiological principles of the biological retina have been adapted to accomplish data sparsity and high dynamic range at the pixel level. These bio-inspired neuromorphic vision sensors alleviate the more serious bottleneck of data redundancy by responding to changes in illumination rather than to illumination itself ······
The analysis of activity is understanding motion pattern and in both the event-based and frame-based data activity recognition tasks, the objective is to decode the most relevant motion embedded in the scene and employ machine learning to perform the analytics ·····
In anomaly detection, the separability between normal and anomaly classes is correlated with the temporal resolution. Dynamic vision sensors alleviate this disadvantage by recording only the pixels which undergo change in intensity, thereby resulting in sparse spatial data and stream of events at microsecond time resolution ·····